An Application of the Tensor-Based Approach to Mortality Modeling

With the increasing availability of temporal data, researchers often analyze information stored in matrices, in which entries are replicated on different occasions. Such multidimensional data can be stored in 3-way arrays or tensors to be analyzed. A collection of 3-way arrays can also be available leading to 4-way arrays. In this work, we apply a tensor-based method, the Tucker4, to mortality data provided by the World Health Organization, referred to 4 dimensions (causes of death, age groups, years, and countries) and organized in a 4-way array. We carry out the analysis on the total population. Our findings reveal some peculiar aspects of the mortality phenomenon.